Where We Left Off
Part 12 ended with a decision taken and a border still secret. This part begins on the morning the secret was published.
Part 12 covered thirty-two years, from 1915 to the summer of 1947. It began with the Ghadar Party, made by Punjabi migrants in North America who discovered that being a British subject meant nothing once they left India. Their rising failed. Their language survived.
It went through the First World War, in which Punjab supplied roughly half of India's combat soldiers while holding about one in thirteen of India's people. It went through 1919 — the Rowlatt Act, the killing at Jallianwala Bagh, and two months of martial law. It went through the Akali morchas of the 1920s, which took the Sikh shrines back from hereditary managers and, almost by accident, built a permanent political machine. It went through Bhagat Singh, hanged at twenty-three.
Then it spent its second half on something that gets left out of most tellings. For twenty years Punjab was not governed by the Congress or by the Muslim League. It was governed by the Unionist Party — Muslim, Hindu and Sikh landholders in one party, organised around the countryside against the towns. In 1937 the Unionists took about 95 of 175 seats. The Muslim League took two.
By 1946 the League had 73 seats and the Unionists 20. The coalition that governed without the League fell on 2 March 1947. Mass killing began in the Rawalpindi division within days — three months before partition was announced. On 23 June the Punjab Assembly, voting in two separate sections, chose to divide the province. A British lawyer who had never been to India was given about five weeks to draw the line.
He finished before independence. The award was not published. On 14 and 15 August 1947, millions of Punjabis in the central districts celebrated or dreaded a freedom whose border they had not been told.
That is where this part opens.
How this book is built#
Six kinds of box run through every part of this series. They are the whole method, so it is worth knowing what each one is for before you meet it.
What it does: explains a hard word the moment it first appears. You never have to hold an unexplained term in your head, and you never have to go hunting in a glossary.
What it does: turns a number you cannot picture into something you can. Ten million people is a figure. Ten million people is also every single person in a large country getting up and walking, in both directions, in about four months.
What it does: shows the evidence itself — the census column, the relief camp register, the commission report, the satellite reading of the water table — and says plainly what it can and cannot prove. This part leans on it heavily, because much of what people state confidently about these years rests on counts that were never properly made.
What it does: sets out a question where serious people disagree. Each side gets a block of its own, with its strongest honest case — not a version built to be knocked down. Then a verdict block says where things stand and what would settle it.
What it does: digs out something a claim quietly takes for granted without proving. This is the signature of the series. It is not a disclaimer and it is not a caveat. It is the unexamined thing sitting underneath an argument that everybody is having.
What it does: closes every chapter with the key points in the plainest words available. If you read only these ten boxes, you would still have the story of this part.
A warning about this part#
Everything in Parts 1 to 12 was finished. The people were dead and the arguments about them were arguments between historians.
This part is different. Some of the people in it are alive. Some of the events are inside living memory, and there are families in Punjab today who lost somebody in them. The 1980s are not settled history; they are a wound that different people describe in incompatible ways, and describing them at all is a political act in India right now.
So I will do here what I have done throughout, only more carefully. Where there is a count, I will give it and say who made it. Where there is no reliable count, I will say so instead of choosing a number that suits a story. Where two accounts cannot both be true, both go in an Argument box at full strength, and I will not resolve it for you.
You may find that unsatisfying in a few places. I think it is the only honest way to write these particular seventy-nine years.
1The Line Made Public
On 17 August 1947 Punjabis were finally told which country they were in. By then a great many of them had already been killed for being in the wrong one.
1.1 — Two days late#
India became independent at midnight on 14 August 1947. Pakistan had been created hours earlier. The border between them, in Punjab, was not made public until 17 August.
Sir Cyril Radcliffe had finished his work before independence. The decision to hold it back was taken at the top. The reasoning was that if the award came out under British authority, the British would have to police the reaction. If it came out after the handover, the new governments would.
So for three days, in the most mixed districts in Punjab, nobody knew.
Imagine being told your country becomes independent on Friday, and being told on Sunday which of two new countries your village is in.
Now add this. Deciding wrongly does not mean inconvenience. It means your family may be killed. And you have to decide on Friday.
People guessed. Some guessed from the religion of the local majority. Some guessed from rumour. In several places the local officials guessed too, because they had not been told either, and raised the wrong flag on 15 August and had to take it down again.
1.2 — What the award actually said#
The line ran roughly north to south through the middle of Punjab. Lahore went to Pakistan. Amritsar went to India.
Two decisions drew the loudest complaints, and still do.
Gurdaspur. The district had a small Muslim majority overall. Three of its four tehsils went to India; only Shakargarh went to Pakistan.
Ferozepur. The Ferozepur and Zira tehsils had Muslim majorities. Both went to India.
Tehsil: an administrative unit below a district. A district contains several tehsils. It matters here because the Boundary Commission's rule — put contiguous majority areas on the side of their majority — gives a different border depending on whether you count by district, by tehsil, or by village. Nobody ever specified which.
Gurdaspur is the single most argued-over stretch of the Radcliffe line, because the only land route from India towards Kashmir at the time ran through it. In October 1947 Kashmir became the subject of a war that has not fully ended.
Gurdaspur had a Muslim majority. Under the stated rule it should have gone to Pakistan. Earlier working maps, including one shown to the parties in early August, apparently placed more of it on the Pakistani side. Mountbatten met Radcliffe privately during the drafting. India's access to Kashmir depended on this award, and India took Kashmir's accession three months later. On this reading the criteria were the cover and the road was the point.
The award's own reasoning rests on the canal and river system: the Upper Bari Doab Canal irrigated land on both sides, and Radcliffe's terms of reference told him to weigh "other factors" as well as population. Batala and Gurdaspur tehsils have substantial non-Muslim populations, and the district majority was thin. Kashmir's accession was not foreseeable in early August — it followed an invasion nobody had scheduled. Both sides denounced the whole award, which is not what you expect from a fix.
What would settle it: a contemporaneous record of Radcliffe's reasoning before and after those meetings. His working papers were destroyed. That is itself a fact worth sitting with.
Why people care so much: because if the line was drawn for a road, then Kashmir — and three wars — began with a pencil, not a principle.
1.3 — What a border cuts#
A line on a map of Punjab does not just separate people. Punjab had been engineered.
Part 11 described the canal colonies: a vast irrigation system built by the British across the western districts, turning dry land into the most productive farmland in India. That system was designed as one machine. The line cut it.
Headworks ended up on one side and the fields they watered on the other. The Madhopur headworks on the Ravi went to India; canals it fed ran into Pakistan. The same was true at Ferozepur on the Sutlej. Within months this became a live weapon: in April 1948 India briefly stopped water to canals running into Pakistan. That single act is the direct ancestor of Chapter Seven.
Roads and railways were cut the same way. Lahore and Amritsar are about fifty kilometres apart and had functioned as one economic region for centuries. Lahore lost its hinterland. Amritsar lost its market and became a border town with a dead frontier on two sides.
Almost every account of 1947 assumes that a border sorts people who were already separate. The task is described as "dividing" — as though Punjab were a bag of two kinds of object that could be tipped into two piles.
That assumption is wrong at the level of the village. In central Punjab, communities did not live in blocks. They lived in the same settlements, in occupations that fitted together — a Muslim weaver, a Hindu trader, a Sikh cultivator, all dependent on each other's work.
So the border did not sort a mixture. It tore a fabric, and then people had to be moved to make the map true after the fact. Notice the order. The map came first. The population was rearranged to fit it.
This matters for a question people still argue about: was the killing an unfortunate side effect of partition, or was it the method by which partition was carried out? If you assume the communities were already separate, the killing looks like an accident. If you look at where they actually lived, it looks like the mechanism.
1.4 — The force that was meant to hold it#
The one preparation made for trouble was the Punjab Boundary Force, set up at the end of July under Major-General Thomas Rees. It had around 50,000 men.
Fifty thousand soldiers sounds large. Spread it over the twelve districts of central Punjab and you get roughly one soldier for every 300 or 400 people, across an area bigger than Belgium, with no helicopters, few radios and a road system that was being blocked by refugee columns.
Put differently: if the trouble started in more than a handful of places at once, the force could not reach it. It started in hundreds.
The force also had a problem inside it. It was made of units of the old Indian Army, which had just been divided between two countries on religious lines. Soldiers were being asked to police violence against and by their own communities, while their own families were on the move. Some units performed with discipline. Some did not, and there are credible accounts of soldiers taking part.
The Punjab Boundary Force was disbanded on 1 September 1947, about five weeks after it was formed, at the exact moment the violence was at its height. Each new government said it would handle its own side.
The border was finished before independence and published two days after it, so that the new governments, not the British, would face the consequences. For three days the most mixed districts in Punjab did not know which country they were in.
Gurdaspur and Ferozepur are still argued about, because Gurdaspur carried the road towards Kashmir. Radcliffe's working papers were destroyed, so the argument cannot be closed.
The line cut a single engineered irrigation system in half, leaving headworks and fields in different countries. That is where Punjab's water disputes begin.
The only force prepared for trouble was disbanded five weeks after it was formed, just as the killing peaked.
2The Killing
Nobody counted. That is not a gap in the record — it is the central fact about the record, and it shapes every claim made since.
2.1 — It did not start in August#
The violence of 1947 is usually dated to August. Part 12 showed that it began in March, in the Rawalpindi division, months before anyone knew there would be a border.
That earlier round matters for understanding what came later. Refugees from Rawalpindi arrived in central and eastern Punjab through the spring, carrying accounts of what had been done to their villages. Some of them were among the men who organised what happened in August, in the other direction.
This is the ugliest mechanism in the whole story and it needs stating plainly. Each round of killing produced witnesses who moved somewhere else, and their arrival supplied both the motive and the manpower for the next round. That is why the violence moved in waves rather than exploding everywhere at once.
2.2 — It was organised#
The common phrase is that Punjab "descended into madness". The evidence does not support the picture that phrase creates.
What the record shows, from police reports, army accounts and survivor testimony, is a pattern of armed bands. They had names, leaders and transport. They were often built around men with military training, of whom Punjab had an enormous supply — hundreds of thousands demobilised after the Second World War. On all sides.
They attacked minority settlements, usually at a numerical advantage of many to one. They frequently attacked at the moment a community was most exposed: a village whose men had left, a column on the road, a train.
Jatha: a band or squad, introduced in Part 12 for the unarmed Akali volunteers of the 1920s. In 1947 the same word was used for armed bands. The word did not change; what it named did.
This is worth noticing, because the vocabulary of the earlier non-violent movement was carried straight into the violence, which is one reason participants could tell themselves it was continuous with something honourable.
Three kinds of evidence, none complete.
Official: police first-information reports, district records, and the daily situation reports of the Punjab Boundary Force. These are contemporaneous, which makes them valuable, but administration in many districts had partly stopped working, and both new governments had reasons to keep numbers down.
Relief: registers from refugee camps and from the Military Evacuation Organisation, recording arrivals, deaths in camp, and people reported missing.
Testimony: a very large body of survivor accounts, most collected decades later. It is rich on what happened in one village and unreliable for totals — people cannot count what they were running from.
What all three agree on is the pattern: organised attacks by armed groups on outnumbered populations. What none of them supports is a reliable total.
2.3 — The trains#
Trains carried refugees in both directions. They were also the easiest targets in Punjab: a known route, a fixed timetable, hundreds of people packed in with nowhere to go.
Trains were stopped, sometimes with a rail removed, sometimes by men waiting at a station. Attacks on trains killed people in numbers that no village attack could match, and they were reported on both sides, which made them uniquely effective at producing the next attack.
2.4 — What happened to women#
A large part of the violence was directed specifically at women, and it needs to be described rather than gestured at.
Women were abducted in very large numbers, and were killed by attackers. Women were also killed by their own families, and killed themselves, to avoid capture — a fact recorded on all sides, and one that families themselves recounted afterwards as honourable.
The two governments later ran a joint operation to recover abducted women, under an agreement in 1947 and then an Indian law in 1949. The working estimates used by that operation were in the region of 50,000 Muslim women taken in India and 33,000 Hindu and Sikh women taken in Pakistan.
Those numbers are estimates made by officials for an administrative purpose, not a count. Treat them as an order of magnitude: tens of thousands on each side.
The recovery operation ran for years. It also took women back from households where they had since had children, sometimes against their stated wishes — so it produced a second displacement of its own.
Both governments treated recovery as obviously right, and both used the same reasoning: a woman taken from one community belonged with that community, and returning her restored something.
The assumption underneath is that a woman's community membership is a fact about her that cannot change, and that her own statement about where she wished to live could be overridden by it.
You can see it in the law's structure: it applied to women, not to men; and a woman's testimony that she had chosen to stay was frequently treated as evidence that she was not free rather than as evidence of choice.
I am not telling you the recovery operation was wrong. Many women were being held by force and wanted out. I am pointing at the premise that made it possible not to ask.
2.5 — The number nobody has#
Published estimates for deaths in the partition of Punjab range from about 200,000 to about two million. That is a tenfold spread, and it has not narrowed in seventy-five years.
Lower estimates, in the low hundreds of thousands, are built from what can actually be documented: recorded burials and cremations, camp registers, official district returns. Their defenders argue that this is the only method that is a count at all, and that higher figures are arrived at by demographic subtraction, which folds in everyone who died of disease, everyone who was never traced, and everyone who simply settled somewhere unrecorded.
Higher estimates, up to two million, are built by comparing the 1941 census with post-partition population accounting and treating the shortfall as deaths. Their defenders argue that documentation collapsed exactly where killing was worst, that whole villages died with nobody left to report them, and that a method which counts only recorded deaths in a period when recording stopped is guaranteed to undercount.
Many historians settle on a range of several hundred thousand to about a million, and treat any single figure as false precision. They note that the demographic method cannot separate the dead from the untraced, and that the documentary method cannot see past the edge of a functioning district office.
What would settle it: nothing available. The 1951 censuses were taken after the movement was complete, and neither state ran a serious inquiry into the dead. That decision — not to look — was made by both governments, and it was not an oversight.
Why people care so much: because the figure is used to assign blame. A larger number in one direction supports one national story; a larger number in the other supports the opposite. Be suspicious of anyone who states a precise total with confidence, in either direction.
The violence began in March 1947, before the border existed, and moved in waves, because each round produced refugees who carried the motive and the manpower into the next.
It was organised, not spontaneous — armed bands, often led by demobilised soldiers, attacking outnumbered populations and trains. On all sides.
Tens of thousands of women were abducted on each side, and a recovery operation later moved many of them again, sometimes against their wishes.
Nobody counted the dead. Estimates run from 200,000 to two million and the gap is a difference of method, not of evidence. Neither government wanted the number known.
3The Walking
In about four months, roughly ten million people crossed Punjab in opposite directions. Almost none of it was planned by anybody.
3.1 — The scale#
Partition produced the largest movement of people in modern history. Across the whole of India and Pakistan the figure usually given is between twelve and eighteen million. The great majority of it happened in Punjab, in the second half of 1947.
Take roughly ten million people. That is more than the population of London. Now make all of them leave, on foot, in both directions, along the same few roads, in about four months, during and after the monsoon.
The largest foot columns — called kafilas — ran to tens of thousands of people. One column is recorded at over 400,000, stretching for many kilometres and taking several days to pass a fixed point.
A column moving at walking pace with bullock carts covers perhaps fifteen kilometres a day. From Lahore to Amritsar took under a week. From the far western districts it could take a month.
Kafila: a caravan or convoy. In 1947 it meant a column of refugees moving on foot with their carts and animals, usually with some armed escort, sometimes military, often not.
Muhajir: Arabic-derived word for migrant, used in Pakistan for those who came from India. It carries a religious echo — the original muhajirun made a migration for faith — which turned a catastrophe into a founding act. India used the plainer word refugee and later displaced person.
3.2 — Nobody planned this#
The most striking administrative fact about 1947 is that no serious scheme existed for population transfer, because officially there was not going to be one.
The working assumption on all sides, stated publicly by Jinnah and by Congress leaders, was that minorities would stay where they were and be protected as citizens. Chapter Ten of Part 12 named this as the hidden assumption of the whole summer.
So there were no reception camps ready, no ration stocks positioned, no transport organised. Everything was improvised after the movement had already started, by administrations that were themselves splitting in two and losing half their staff to the other country.
Refugee movement is far better documented than the deaths, for a mundane reason: people who arrive have to be fed.
Camps kept ration registers. The Military Evacuation Organisation logged convoys and escort numbers. Land resettlement, which is the subject of the next chapter, required every arriving family to file a claim stating what they had left behind — producing millions of individual records that still exist.
This creates a lopsided archive, and it is worth naming. We know a great deal about the people who arrived and very little about the people who did not. The record was built by the act of surviving.
3.3 — The camps#
Camps appeared along the routes and around the towns. The largest, at Kurukshetra, held around 300,000 people at its peak — a temporary city with no sanitation built for it, in the middle of a cholera season.
The immediate killers in camps were the ordinary ones: cholera, dysentery, exposure, and hunger. Nobody knows how many died this way either, and these deaths sit inside the disputed totals from the last chapter.
3.4 — What was actually lost#
People carried what could be carried. What they left was everything else, and it is worth being concrete about what "everything else" means.
It means standing crops. It means houses, wells, workshops, tools, livestock, and the shop's ledger of who owed what. It means the network of people who bought your work — which is the part that cannot be reissued by any government.
Two cities lost their function overnight. Lahore, the capital of Punjab for centuries and its centre of publishing, law and education, kept its buildings and lost the population that had run half of them. Amritsar kept its shrine and lost its market.
Both states described what happened as an exchange of populations. The word is doing heavy work.
An exchange implies two comparable movements that cancel out. It suggests a Muslim family leaving Ludhiana and a Sikh family arriving from Lyallpur are, in the ledger, the same event.
They were not. The land left in east Punjab and the land left in west Punjab were of different quality, different acreage and different value — the western canal colonies were the best-irrigated land in the subcontinent. The urban property differed. The occupations differed.
The word "exchange" also does something quieter. It turns people who were driven out into participants in a transaction, which makes the movement sound agreed. Nobody agreed to it. The next chapter shows what happened when a government had to do the arithmetic that the word "exchange" pretends is already done.
Roughly ten million people crossed Punjab in about four months, in both directions, mostly on foot. It is the largest movement of people in modern history.
None of it was planned, because every leadership had publicly assumed minorities would stay. Camps, rations and escorts were improvised after the walking had started.
The archive is lopsided: we know about arrivals because arrivals had to be fed. The record was built by surviving.
Calling it an "exchange" hides the fact that what each side left behind was not equivalent — and that nobody agreed to any of it.
4Building a Province From Half a Province
East Punjab had to settle millions of farmers on land that was smaller and worse than the land they had lost. What it invented to do that was one of the most remarkable pieces of administration in modern India.
4.1 — The arithmetic that would not work#
East Punjab received around four and a half million refugees. Most were farming families. Every one of them had left land behind.
The land available to give them was the land left by Muslims who had gone the other way. Two problems appeared immediately.
First, there was less of it. Claims filed by arriving families came to substantially more acreage than existed on the Indian side to hand out.
Second, what existed was worse. The best land in undivided Punjab was in the western canal colonies. That was now in Pakistan.
Picture a village where forty families arrive, each holding a paper saying they used to own ten acres of well-irrigated land. In the village there are thirty farms of six acres, most of them rain-fed.
You cannot satisfy the papers. You can only decide, in public, how to disappoint everybody in a way they will accept without a riot — while they are exhausted, armed, and have recently seen what people do.
4.2 — The standard acre and the graded cut#
The scheme designed to solve this is associated with the officer Tarlok Singh, and it had two ideas in it.
The first was the standard acre. Instead of measuring land by area, the scheme measured it by productivity. A fixed yield of produce became the unit. So an acre of top canal-irrigated land in Lyallpur counted as more standard acres than an acre of dry land in Hisar. A claim was converted into standard acres, and an allotment was made in standard acres.
Standard acre: a unit of land measured by what it produces rather than how big it is. Defined as the area yielding a fixed quantity of grain. Ten poor acres might equal four standard acres; four excellent acres might equal six.
Why it matters: it let one office compare a claim about land in a country it no longer controlled with a farm it could actually hand over.
The second idea was the graded cut. Everyone got less than they claimed, but not by the same proportion. Small claims were cut lightly. Large claims were cut heavily — the largest by a very large fraction.
This was deliberate and it was redistributive. A displaced smallholder came out close to whole. A displaced large landlord did not.
Indian resettlement of Punjab refugees is often called one of the great administrative achievements of the twentieth century. It is also called a long injustice. Both descriptions come from serious people.
Millions of destitute families were given land, in a few years, by a new state with no money, using records that had to be reconstructed from claims. There was no mass permanent camp population — unlike several other twentieth-century refugee situations that are unresolved today. Agriculture in east Punjab recovered quickly, which set up everything in Chapter Six. And the graded cut meant the burden fell hardest on those with the most, which is not how such schemes usually go.
The scheme dealt in land claims, so it served people who had owned land. Landless labourers, artisans, weavers and menial workers — a very large share of the refugee population, and disproportionately Dalit — had no claim to file and therefore received little or nothing. Urban refugees were handled far worse than rural ones. Verification depended on documents and on local witnesses, which favoured the connected. And large numbers of claims were inflated, which everyone knew.
What would settle it: a systematic study of what happened to non-claimant refugees. It has never been done at scale, partly because the archive — as Chapter Three noted — was built out of claims, so the people without claims are the ones it cannot see.
4.3 — A state built out of leftovers#
East Punjab in 1947 was not a tidy half. Alongside the British province there had been princely states — Patiala, Nabha, Jind, Kapurthala, Faridkot and others — which were not part of British Punjab at all.
In 1948 eight of them were merged into a single unit: the Patiala and East Punjab States Union, PEPSU, with its capital at Patiala. It lasted until 1956, when it was absorbed into Punjab.
Princely state: a territory ruled by an Indian prince under British paramountcy rather than governed directly by British officials. Punjab had many. Their laws, taxes and police were their own.
PEPSU: Patiala and East Punjab States Union, 1948–1956. Worth remembering because Punjab's borders changed four times in nineteen years — 1947, 1956, 1966, and administratively again with Chandigarh — which is one reason its politics has been unusually preoccupied with territory.
4.4 — A capital with no history#
Lahore was gone. East Punjab had no capital city, and it declined to promote an existing town.
Instead it built one. Chandigarh was planned from scratch on farmland at the foot of the Shivaliks. After the American planner originally engaged withdrew following his partner's death, the work passed to the Swiss-French architect Le Corbusier, with a team including Pierre Jeanneret, Maxwell Fry and Jane Drew. Construction began in the early 1950s.
It is a grid of numbered sectors, each intended to be self-sufficient, with wide roads and heavy concrete public buildings.
Chandigarh was built on an assumption that was never argued for in public, because it did not feel like an assumption at the time: that a city is a machine that can be designed in advance, and that a good design will produce a good social life.
Notice what this rules out. Old Punjabi towns grew around a market, a shrine, a river crossing or a fort. They were dense, mixed, and chaotic, and their form recorded their history. A planned city has no history to record, so it has to substitute intention.
Chandigarh has among the highest incomes and best civic amenities in India, so the assumption was not simply wrong. But the city is also frequently described as lacking street life, and its plan has proved poor at absorbing the unplanned — the informal settlements it did not budget for have grown around it rather than in it.
The deeper point applies well beyond one city, and it returns in Chapter Six: a design optimised for one measurable good tends to be brittle against the things nobody measured.
There is one more thing about Chandigarh. When Punjab was divided again in 1966, both new states claimed it. It was made a union territory serving as capital to both — a temporary arrangement, still in force sixty years later, and still an active grievance. A city built to give Punjab a fresh start became a permanent dispute.
East Punjab had to settle about four and a half million refugees on land that was smaller in area and poorer in quality than what they had left.
It invented the standard acre — land measured by yield, not size — and the graded cut, which took a much larger share from big claims than small ones. It worked, and it was genuinely redistributive.
But it was built on land claims. The landless, the artisans and the urban poor had nothing to claim and got very little. The archive cannot see them, because the archive is made of claims.
Punjab lost its capital and built one from nothing. Then it was divided again, and the new capital became something both halves are still arguing over.
5The Language Question
For nineteen years Punjab argued about which language it spoke. What was actually being argued about was who would hold a majority — and both sides knew it.
5.1 — What was being asked for#
After 1947, states across India were being reorganised so that each major language had a state where it was the official language. Andhra, Karnataka, Maharashtra and others were made or remade on that principle.
The Akali Dal asked for the same thing for Punjabi: a Punjabi-speaking state, a Punjabi Suba, carved out of the larger Punjab that then included the Hindi-speaking south and east.
Suba: province. Punjabi Suba: the demand for a state in which Punjabi was the language of government, courts and schools.
Gurmukhi and Devanagari: two scripts. Punjabi in India is written in Gurmukhi, the script of the Guru Granth Sahib. Hindi is written in Devanagari. Because the script is tied to a scripture, a fight about script is never only about writing.
The demand looked like the ones being granted elsewhere. It was refused for nineteen years.
5.2 — The census answer#
The reason given for refusal was that Punjab was not simply a Punjabi-speaking region. The evidence given was the census.
And the census said something extraordinary. In 1951, and again in 1961, very large numbers of Punjabi Hindus recorded their mother tongue as Hindi.
This is one of the clearest cases in the whole series of a source measuring something other than what it appears to measure.
The census asks for mother tongue and records the answer. It does not verify it. In Punjab in 1951 and 1961 the answer had become a political declaration: organised campaigns, associated with the Arya Samaj and sections of the Hindu press, urged Punjabi Hindus to return Hindi, precisely because the census would be used to decide whether a Punjabi state was justified.
So the census figures are excellent evidence of what people declared and poor evidence of what people spoke. Many of those returning Hindi spoke Punjabi at home, in the street and at work.
Keep this in mind whenever anyone cites census language data in South Asia. It records an answer, and an answer can be a vote.
Both sides in this fight accepted a premise without examining it: that one person has one mother tongue, and that a district can therefore be assigned to one language.
In Punjab this was never true. A single household could use Punjabi at home, Hindi or Urdu for correspondence, and Sanskrit or Gurmukhi for prayer. Before 1947 Urdu had been the language of the courts and much of the press across the whole province.
The census form has one box. The reality has several. Once the state requires one answer, people supply one — and then the state treats its own form as a discovery about the population.
Watch this move. It is exactly the same one as the Sikh electoral roll in Part 12, chapter five: an administrative need for a hard line is imposed on soft ground, and the line then starts to look like a description of reality.
5.3 — Why it took nineteen years#
The States Reorganisation Commission reported in 1955 and refused the Punjabi Suba, while granting linguistic states elsewhere.
The stated reason was that Punjabi and Hindi were not sufficiently distinct in the region and that the demand lacked general support. The reason most historians point to is different: a Punjabi-speaking state, drawn as proposed, would have a Sikh majority. Every other linguistic state being created had a religious composition nobody had to think about. This one did not.
That is why the demand attracted a suspicion no other state movement faced — that language was a cover for a religious state. Akali leaders denied it, pointing out that they were asking for exactly the principle India had already adopted. Opponents replied that the principle would not have been so attractive if the arithmetic had gone the other way.
Punjabi is a distinct language with its own literature going back to Baba Farid, as Part 6 showed. India had adopted linguistic reorganisation as national policy and applied it everywhere else. Refusing it in exactly one place, where the language happened to be associated with a minority faith, is the discrimination — not the demand. Akali leaders consistently framed it in the language of parity, and the movement drew Punjabi-speaking support beyond Sikhs.
The demand was led by an explicitly Sikh party, run from institutions controlled by the SGPC, and mobilised through gurdwaras. Its most committed supporters spoke of it as securing a homeland where Sikhs would not be a permanent minority. The script chosen was the scripture's script. And the strongest evidence is that when the state was finally granted, the boundary drawn was the one that produced the Sikh majority.
What would settle it: nothing, because the question assumes the two motives can be prised apart in the heads of people who did not experience them as separate.
Why people care so much: because if it was a language demand, then nineteen years of refusal was discrimination; and if it was a religious demand, then the refusal was principled. The label decides who was wronged.
5.4 — 1966#
The state was granted in 1966, after the war with Pakistan in which Punjab had again been the front line, and following a long campaign that included fasts, mass arrests and one death by self-immolation.
The Punjab Reorganisation Act took effect on 1 November 1966. It produced three things: a smaller Punjab with a Punjabi-speaking, Sikh-majority population; a new state of Haryana in the Hindi-speaking south-east; and hill districts transferred to Himachal Pradesh.
Chandigarh, sitting on the boundary, was made a union territory and shared as capital by Punjab and Haryana. The river waters were not divided along with the land, but administered jointly. Both of those decisions were described as temporary.
Neither has been resolved. Chapter Seven is about the second one.
India reorganised its states by language everywhere. Punjab asked for the same and was refused for nineteen years.
The refusal rested on census figures in which large numbers of Punjabi Hindus declared Hindi as their mother tongue — often at the urging of organised campaigns, precisely because the census would decide the question. The census recorded an answer, not a language.
The demand was both a language demand and a community demand, and in Punjab those two things had been growing into each other since the colonial census.
Punjab got its state in 1966, minus Haryana and the hills — and with its capital and its river water left unresolved. Both are still unresolved today.
6The Green Revolution
Punjab was asked to feed India, and it did. The bill arrived thirty years later, and it is still being paid.
6.1 — The problem it solved#
In the mid-1960s India could not feed itself. Two failed monsoons and a war left the country dependent on imported American wheat, shipment by shipment, in a way that was politically humiliating and strategically dangerous.
The response was a package. Not a single invention — a package, and the package matters more than any part of it.
New seed varieties of wheat, developed in Mexico under Norman Borlaug and adapted at Indian institutions including Punjab Agricultural University at Ludhiana. Chemical fertiliser. Assured irrigation. Guaranteed purchase by the government at an announced price. Rural credit to pay for it all.
High-yielding variety (HYV): a plant bred to produce far more grain per plant. The key trick was making the plant shorter. Tall wheat given lots of fertiliser grows tall and falls over. Short, stiff-stemmed wheat puts the extra growth into grain instead of stalk and stays standing.
Minimum Support Price (MSP): a price announced by the government before sowing, at which it undertakes to buy the crop. It removes price risk. A farmer deciding what to plant knows what it will fetch.
Procurement: the government actually buying that grain, through agencies, and storing it for the public distribution system.
Punjab was chosen to lead because it already had what the package required: canal irrigation from the colonial period, consolidated landholdings, a literate farming population, and — after Chapter Four — a structure of owner-cultivators rather than absentee landlords.
6.2 — What happened#
Punjab is about one and a half per cent of India's land area.
Through the peak decades it supplied, in some years, somewhere between a half and two thirds of all the wheat the central government bought for the national food system, and a very large share of the rice.
Put it in kitchen terms. For decades, a large fraction of the subsidised grain distributed to poor households from Kerala to Assam was grown on a patch of ground smaller than the smallest of several Indian states.
Wheat yields multiplied. Punjab became the highest-income state in India by the 1970s and early 1980s. Villages got electricity, roads, tractors and pucca houses on a scale unmatched elsewhere in the country. India stopped importing food.
By any measure the policy set out to achieve, it succeeded completely.
6.3 — Who gained#
Not everyone gained equally, and the pattern of who gained shaped the next twenty years of Punjabi politics.
The package rewarded those who could buy inputs and wait for a harvest. Owner-cultivators with medium and large holdings did best. Farmers with tiny holdings gained less, because the fixed costs of a tubewell and a tractor do not shrink with your acreage.
Agricultural labourers gained real wages, especially at harvest, and Punjab drew in large seasonal migration from Bihar and eastern Uttar Pradesh. But labourers did not gain assets, and the gap between those who owned land and those who worked it widened in absolute terms even as everyone's income rose.
The entire package assumed that what a farmer maximises is yield per acre this season. Every instrument pointed that way: the seed, the price guarantee, the credit, the extension advice, the subsidy on power and fertiliser.
Nothing in the package priced the water, the soil or the next generation's options. Not because anyone decided those did not matter, but because they were not in the equation being solved. The equation being solved was: India is hungry now.
Part 1, chapter 8, raised a gap that this series said would return: the gap between what is good for the group and what is good for the person. Here it is, in its most consequential Punjabi form, and it runs in an unexpected direction.
For each individual farmer, sinking a deeper tubewell and growing rice is rational. It is the highest-return use of his land under the prices he faces. For all farmers together, every deeper tubewell lowers the water table for everyone, including him.
Nobody is behaving badly. Every actor is doing the sensible thing. That is precisely what makes it hard to stop — and it is why Chapter Seven is not really a chapter about engineering.
6.4 — The bill#
Four costs came due, all of them slowly enough that no single year looked like a crisis.
Water. Rice is not a traditional Punjabi crop. It needs standing water in a state with a semi-arid climate. The water came from the ground, pumped by electricity that was made effectively free from the late 1990s. Chapter Seven takes this up in full.
Soil. Continuous wheat-rice cycling, with heavy fertiliser and little rotation, degraded soil organic matter and produced micronutrient deficiencies that then required more inputs to correct.
Cost. Input prices rose faster over time than the support price. The margin per acre narrowed, so the response was to farm more intensively, which raised input costs again. Farm debt climbed, and rural indebtedness became one of the defining features of Punjabi life from the 1990s.
Lock-in. This is the worst one. The whole apparatus — assured purchase, the mandi system, credit, machinery, the labour calendar — was built around two crops. A farmer who wants to switch faces the fact that only wheat and rice have a guaranteed buyer.
It ended famine risk for a country of hundreds of millions and removed India's dependence on foreign food aid, which is not a small thing to weigh against soil chemistry. It made Punjabi villages richer than any others in India for a generation, and that wealth bought schools, health care and migration opportunities whose benefits are still compounding in Punjabi families today. Judging a policy that fed a nation by the problems success later produced is judging it against a world where the alternative was not tried — and the alternative was hunger.
It converted a renewable advantage into a depleting one. Punjab traded groundwater accumulated over millennia for three decades of income, and the water is not coming back on any human timescale. It locked the state into two crops it cannot easily leave, produced a debt crisis and a suicide crisis among farmers, and left Punjab's per capita income ranking falling steadily from first place. A gain that consumes the asset that produced it is not development; it is liquidation.
What would settle it: the closest thing to a test is happening now, in states that adopted the same seeds without the same subsidy structure. Their outcomes differ, which is evidence that the design, not the seed, did much of the damage.
Why people care so much: because a farmer in Punjab today is being asked to bear the cost of a policy his father was praised as a patriot for adopting.
The Green Revolution was a package, not a seed: short-stemmed high-yield wheat, fertiliser, assured irrigation, cheap credit, and a guaranteed government price. Punjab led it because it already had the irrigation and the owner-cultivators.
It worked completely on its own terms. Punjab — about 1.5 per cent of India's land — fed a large share of the national grain system and became the richest state in India.
Every instrument in the package rewarded yield this season. Nothing priced the water, the soil, or the next generation.
Each farmer sinking a deeper tubewell is behaving rationally. All of them doing it empties the aquifer under all of them. That is why it is so hard to stop.
7Water
Part 1 promised that Punjab's water disputes would keep coming back. This is where they arrive all at once — between two countries, between three states, and between a farmer and the ground under his own field.
7.1 — Between two countries#
Partition left the irrigation system of Punjab cut in half, with headworks in India and command areas in Pakistan. In April 1948 India briefly stopped water flowing into canals feeding Pakistani land. Water had been demonstrated as a weapon within eight months of independence.
It took twelve years to settle. The Indus Waters Treaty was signed in 1960, brokered by the World Bank.
Its solution was blunt. Rather than share each river, it divided them. India got the three eastern rivers — the Ravi, the Beas and the Sutlej. Pakistan got the three western ones — the Indus, the Jhelum and the Chenab. Each side could largely do as it liked with its own, subject to limits on what India could do upstream on the western rivers.
Of the total water in the six rivers, the treaty gave Pakistan roughly eighty per cent and India roughly twenty. That sounds lopsided until you notice the geography: Pakistan's entire agriculture depends on this one system, while India has other river basins.
For Punjab specifically the effect was different and sharper. The three rivers India kept are the three that run through Indian Punjab. So the treaty concentrated India's whole share into the state that had just been reduced to half its size — and then, six years later, that state was divided again.
The treaty has survived three wars and repeated crises. It is routinely described as one of the most durable water-sharing agreements in the world. It achieved that durability by dividing rather than sharing, which is a real lesson and not an entirely comfortable one.
7.2 — Between three states#
Chapter Five noted that when Punjab was divided in 1966, the land was divided but the river waters were not. They were placed under central administration and allocated by a series of decisions and tribunals over the following decades.
The dispute that grew out of this concerns the Sutlej-Yamuna Link canal, usually just called the SYL.
SYL (Sutlej-Yamuna Link): a proposed canal to carry water from the Sutlej system across Punjab to Haryana, giving Haryana access to the share allocated to it after the 1966 division.
Riparian principle: the idea in water law that the states a river actually flows through have first claim on it. It is central to Punjab's case and contested by Haryana's.
An agreement in 1981 fixed allocations and the canal was begun in 1982. It was never completed. Work stopped during the violence of the 1980s, including the killing of engineers on the project. In 2004 the Punjab legislature passed a law terminating its water-sharing agreements. In 2016 the Supreme Court held that law to be unconstitutional. The canal is still not built, and the matter is still live.
Rivers belong first to the states they flow through, which is the standard riparian principle used almost everywhere else. Haryana is not a riparian state for the Sutlej. Punjab's own water table is collapsing, so transferring water out is not sharing a surplus — it is taking from a state already in deficit. The original allocations were made when Punjab's groundwater was healthy and its cropping pattern was different; applying 1970s numbers to 2020s hydrology is applying a settlement to a world that no longer exists.
Haryana was part of Punjab until 1966 and contributed to the infrastructure that captures this water. A division of a state divides its assets, and water is an asset. Haryana's allocation was agreed, was confirmed by tribunals, and has been upheld in court. Punjab has simply refused to comply for forty years, which makes the dispute a question of whether a state can ignore judicial decisions it dislikes. And Punjab's groundwater crisis is largely the product of growing rice, a choice Punjab made and continues to make.
What would settle it: honestly, a change in what Punjab grows. Nearly every serious analysis concludes that the volume in dispute is smaller than the volume that would be freed by moving a fraction of the rice area to less thirsty crops — which requires assured purchase for those crops, which returns you to Chapter Six.
Why people care so much: because in Punjab this is not experienced as an accounting dispute. It is experienced as a state that was cut down twice being asked to give up one more thing.
7.3 — Between a farmer and the ground#
The dispute that will decide Punjab's future is not with Pakistan or Haryana. It is underground.
Punjab has well over a million tubewells. They run on electricity that has been supplied to agriculture free or nearly free since the late 1990s. They pump water from the aquifer under the state to grow a crop that needs standing water in a place with about half the rainfall it requires.
Aquifer: a layer of underground rock or sand holding water in the spaces between its grains. Think of a very deep sponge under the fields.
Recharge: water soaking down from rain and rivers to refill the sponge. Over-exploited is the official term for a block where more is pumped out each year than soaks back in.
Three independent lines of evidence point the same way, which is what makes this one of the few things in this part that is not contested.
Well measurements. India's Central Ground Water Board measures water levels in observation wells and classifies each block. In Punjab the large majority of assessment blocks are classed as over-exploited — a figure that has risen decade on decade.
Pump depth. Farmers' own equipment records the trend. Centrifugal pumps that worked at shallow depths have been replaced across much of central Punjab by submersible pumps set far deeper, at considerable cost. Nobody spends that money for fun.
Satellite gravity data. Since 2002, satellites measuring tiny changes in Earth's gravity have tracked mass loss from north-west India. Losing water from underground makes the ground very slightly less massive. This method knows nothing about Indian politics or crops, and it agrees with the wells.
When a farmer's wallet, a government meter and a satellite all say the same thing, the thing is true.
Growing one kilogram of rice by flooding takes thousands of litres of water. Punjab produces many millions of tonnes of it a year.
Here is the part that makes people stop. Most of that rice does not stay in Punjab. It goes into the national food system. So Punjab is, in effect, exporting its groundwater to the rest of India — in the form of grain — and getting paid the price of the grain, not the price of the water.
The water is not on the invoice. It has never been on the invoice.
Every arrangement in this chapter — the treaty, the tribunals, the canal, the free power — assumes that water is a quantity to be divided. The disputes are about shares.
That assumption fits a river. A river arrives each year whether or not you argue about it, so the only question is who gets what proportion.
It does not fit an aquifer. An aquifer is a stock, not a flow — a fixed inheritance that took thousands of years to fill. You cannot divide it into annual shares, because it is not producing annual anything. You can only decide how fast to spend it.
So Punjab has spent sixty years in a loud argument about surface water shares, conducted in courts and cabinet rooms, while the resource that actually underwrites its agriculture has been drained without any argument at all — because nobody owns it, nobody meters it, and no tribunal has jurisdiction over it.
The argument everyone is having is not about the thing that will decide the outcome.
The Indus Waters Treaty of 1960 settled the dispute with Pakistan by dividing the rivers rather than sharing them — three east to India, three west to Pakistan. It has survived three wars.
When Punjab was divided in 1966 the land was split and the water was not. The Sutlej-Yamuna Link canal has been unbuilt since 1982. Haryana keeps winning in court; the canal keeps not existing.
The decisive dispute is neither of these. Punjab has over a million tubewells running on free power, pumping an aquifer that took millennia to fill, to grow a crop that needs standing water in a semi-arid state.
Punjab exports its groundwater as rice and is paid for the rice. The water has never been on the invoice.
8The Decade That Is Still Being Argued About
This is the hardest chapter in this series to write. I am going to set out what is documented, name what is disputed, and refuse to hand you a verdict.
8.1 — How I am going to do this#
Everything in this chapter is inside living memory. People who took part are alive. People who lost family are alive. In India today, the words you choose here are read as a declaration of which side you are on.
So I will do three things and no more. State the sequence, which is largely agreed. Give the figures with their sources and their disputes attached. Put the genuine disagreements into Argument boxes at full strength.
What I will not do is tell you who was right. If you want that, this is the wrong book, and I would rather lose you here than pretend to an authority I do not have over events this raw.
8.2 — What was being demanded#
The Akali Dal adopted the Anandpur Sahib Resolution in 1973, revising it in 1978. It is the document at the centre of the decade, and almost nobody arguing about it has read it.
Its contents, in plain terms: more powers for states and fewer for the centre, with the centre restricted to defence, foreign affairs, currency and communications; Chandigarh transferred to Punjab; the Punjabi-speaking areas left out in 1966 transferred in; a larger share of river water; a higher share of army recruitment; and recognition of Sikhism's distinct status, including a demand that it not be treated as a branch of Hinduism in law.
This is a dispute about a text, which makes it unusually tractable — and it was never resolved anyway.
On this reading it is a demand for genuine federalism, of a kind argued for by regional parties in Tamil Nadu, West Bengal and elsewhere. It nowhere asks for a separate country. Its language about state powers closely tracks mainstream federalist argument of the period. Several later commissions on centre-state relations took up similar themes. Read plainly, it is a bargaining document about autonomy.
On this reading, a document leaving the centre only four subjects is asking for a state in all but name, and its religious framing gives it a character no Tamil or Bengali autonomy demand had. Its defenders in the 1980s included people openly calling for separation, and the resolution existed in several versions with contested wording, which allowed different audiences to be told different things.
What would settle it: agreement on which version is authoritative — which never happened, and by the mid-1980s the argument was no longer about the document.
8.3 — The sequence#
In 1982 the Akali Dal launched the Dharam Yudh Morcha, a mass civil-disobedience campaign in the pattern described in Part 12 — jathas courting arrest, in very large numbers.
Alongside it, a separate and increasingly armed current developed, associated with the preacher Jarnail Singh Bhindranwale. Targeted killings began. Through 1983 the state's authority eroded. President's rule was imposed in October 1983.
Bhindranwale and armed followers took up residence in the Golden Temple complex at Amritsar, fortifying buildings inside it, including the Akal Takht.
Akal Takht: "throne of the Timeless One", the seat of temporal authority in Sikhism, founded by Guru Hargobind and described in Part 8. It faces the Harmandir Sahib — the Golden Temple — across the pool.
Its meaning is why what happened there is not comparable to a siege of an ordinary building. It is the institution from which the community's collective decisions are announced.
In June 1984 the Indian Army was sent in. Operation Blue Star ran from 1 to 8 June. It used tanks. Bhindranwale was killed. The Akal Takht was extensively destroyed.
The operation took place in the first week of June, when the complex was crowded with pilgrims for the martyrdom anniversary of Guru Arjan.
The government White Paper gave figures in the region of 490 civilians and militants killed, with 83 soldiers dead. Its defenders note it is the only account based on a systematic tally, produced within months, and that army casualty figures — which are high — are not the numbers a government inflates.
Sikh organisations, several journalists and some later inquiries put the toll far higher, in the thousands. Their case: the complex was full of pilgrims for the anniversary; there was no independent access, since the press was excluded from Punjab and communications were cut before the operation; and bodies were cremated in bulk without individual identification. A count made only by the party conducting the operation, with all observers removed, is not independent evidence.
What would settle it: nothing now. The exclusion of the press and the manner of the cremations removed the evidence at the time, and that has itself become part of the grievance.
On 31 October 1984 Indira Gandhi was assassinated by two of her bodyguards, both Sikhs.
In the days that followed, organised mobs attacked Sikhs, above all in Delhi but also in other cities. Households were identified, often from electoral rolls. Men were killed in the street and burned. Official figures put the Delhi dead at around 2,700; other counts run higher, and there were further deaths elsewhere.
November 1984 is documented far better than June 1984, and the difference is instructive.
It happened in the capital, in front of journalists and diplomats. Citizens' groups produced detailed reports within weeks, naming localities, times and in some cases individuals. Nine official inquiries followed over three decades, the fullest being the Nanavati Commission which reported in 2005.
Those inquiries established two things that are not seriously contested: that the killing was organised rather than spontaneous, with mobs transported and provided with lists and fuel; and that police in many areas did not intervene.
What has not followed is much prosecution. A handful of convictions have come decades later. The gap between what is documented and what has been prosecuted is, for many Sikh families, the whole point.
The word almost every account reaches for is riot. Riot implies a crowd that forms, loses control, and disperses. It implies nobody planned it.
Both June and November 1984 are routinely described using words that carry assumptions of that kind — "riots", "communal frenzy", "the troubles". The same vocabulary was used for 1947 and Chapter Two showed what it hides there.
The evidence in November 1984 shows lists, transport, fuel and a pattern of police inaction. Those are the marks of organisation, not of frenzy. Many Sikh organisations therefore use the word pogrom, and several Indian courts have since used language closer to organised violence than to rioting.
I am not asking you to adopt a particular word. I am pointing out that the most common word already contains a conclusion about whether anybody decided.
8.4 — The years after#
An accord was signed in July 1985 between Rajiv Gandhi and the Akali leader Harchand Singh Longowal, promising Chandigarh to Punjab and referring the other disputes to commissions. Longowal was assassinated the following month. Chandigarh was not transferred.
What followed was roughly a decade of armed militancy and counter-insurgency. Militant groups killed civilians, officials and each other; bus massacres and bombings killed large numbers of ordinary Punjabis of every community. The state responded with emergency laws, mass detention, and a police campaign that became notorious for extrajudicial killings and disappearances.
The disappearances are documented largely because of one man's work. The human-rights activist Jaswant Singh Khalra examined municipal cremation records in Amritsar district and found large numbers of bodies cremated by police as "unidentified", including people whose families knew exactly who they were.
He was abducted in September 1995 and killed; police officials were later convicted of it. Following his findings, a CBI investigation ordered by the Supreme Court identified over two thousand such cremations in that one district alone.
One district. The scale statewide has never been established, because no comparable exercise was carried out elsewhere.
Elections were held in 1992 with a turnout of around a quarter of voters, most parties boycotting. By the mid-1990s the militancy was over. Normal politics resumed in 1997.
On this account the credit belongs to a hard, centralised police campaign that broke the groups' command structures, denied them safe houses, and made membership fatal. Supporters point to how quickly violence fell once the strategy was applied, and argue that a state facing armed groups killing bus passengers had no gentler instrument available.
On this account the movement destroyed its own base. As groups fragmented, they increasingly extorted, killed rivals and killed ordinary Punjabis, including in villages that had sheltered them. Support drained away, and the police campaign succeeded because it was operating in a population that had stopped protecting anyone. Supporters note that repression alone rarely ends insurgencies and often feeds them.
A third position holds that after a decade the population simply wanted it to stop, at almost any price — and that the acceptance of the methods used is itself evidence of exhaustion rather than approval.
What would settle it: a full public accounting of what was done during those years — which has not happened, and its absence is why the argument stays live.
The Anandpur Sahib Resolution read plainly is a demand for federal autonomy, not separation — but the movement around it came to contain people who wanted separation, and both readings had real support.
June 1984: the army entered the Golden Temple complex, killed Bhindranwale and destroyed the Akal Takht, during a crowded anniversary. Casualty figures range from several hundred to thousands and no independent count exists.
November 1984: after Indira Gandhi's assassination, organised mobs killed Sikhs in Delhi and elsewhere — around 2,700 by official count, more by others. Inquiries established organisation and police inaction. Prosecutions have been very few.
A decade of militancy and counter-insurgency followed, with mass killing by armed groups and mass disappearances by the state. Over two thousand secret cremations were confirmed in one district alone.
Almost nothing in this chapter has been settled by a court, a commission or a count. That is not a gap in this book. It is the situation.
9After
Punjab came out of the 1980s peaceful and stopped being rich. This chapter is about the slow problems — the ones with no single event to date them to.
9.1 — The state that stopped growing#
In the 1970s and early 1980s Punjab had the highest per capita income of any Indian state. It has since fallen a long way down the table, overtaken by states that were far behind it.
Nothing dramatic caused this. Punjab did not shrink. It grew slowly while others grew fast.
The reason is structural, and Chapter Six laid the ground for it. Punjab's prosperity was built on agriculture, and agriculture cannot go on doubling. Meanwhile the growth in India after 1991 came from manufacturing and services, which concentrated in states with ports, large cities and industrial policy.
Think of two families. One owns a very good farm; the other opens a workshop. For thirty years the farm family is richer. Then the workshop family's business grows in a way a farm cannot, because you can add a second workshop and you cannot add a second monsoon.
Punjab is the farm family. It did nothing wrong. It ran out of the kind of growth it had.
Punjab's small industrial base — Ludhiana's hosiery and bicycle parts, Jalandhar's sports goods, Batala's machine tools — did not scale as India's economy opened. A landlocked border state with a decade of insurgency behind it was not where investment went in the 1990s.
9.2 — Debt, drugs and departure#
Three things get named whenever Punjab's present is discussed, and they are connected.
Debt. The squeeze described in Chapter Six — input costs rising faster than the support price — produced sustained rural indebtedness. Farm suicides became a recognised phenomenon from the 1990s, concentrated among small farmers and agricultural labourers.
Drugs. Punjab has a serious substance problem, involving opioids in particular. It sits on a smuggling route, it has money, and it has large numbers of young men with time.
Be careful with claims about Punjab and drugs. The subject attracts very large round numbers, often quoted as if from a survey — "seventy per cent of Punjabi youth" and similar. That figure and its relatives have no reliable source.
What does exist: household surveys of substance use, treatment-centre admission data, and a 2015 study for the central government which estimated the number of opioid-dependent people in Punjab in the low hundreds of thousands. That is a serious public-health problem in a state of about thirty million. It is not most of the young population.
The inflated version is worth noticing as a phenomenon in itself. It is used politically by everyone, and it makes the actual problem harder to treat by turning a medical condition into a stereotype about a people.
Departure. Emigration from Punjab is old — Part 12 opened with Punjabis in Oregon in 1913. What is new is the scale and the direction of the aspiration. Leaving is now a mainstream plan rather than an adventure, and the money involved has produced an entire industry of agents, language schools and study-visa routes, some legitimate and some not.
9.3 — The thing that gets left out#
Part 3 of this series made a promise. It said that Punjab's unusually weak Brahmin presence, and the rejection of caste by the Sant tradition and by the Sikh Gurus, raised a question to be taken up later: did it work?
Here is the answer, and it is uncomfortable.
Punjab has the highest proportion of Scheduled Caste people of any Indian state — around 32 per cent in the 2011 census. Roughly one in three Punjabis.
Land ownership runs the other way. Dalits own only a very small share of Punjab's agricultural land — a few per cent — despite being a third of the population. The great majority of Punjab's farmland is owned by the intermediate cultivating castes, above all Jat Sikhs.
So the state with India's largest proportion of Dalits is also the state where the gap between population share and land share is at its widest.
Chapter Four explained part of the mechanism. Resettlement worked through land claims, and people without land had nothing to claim. The scheme did not discriminate on paper. It did not have to; it simply distributed on the basis of a thing the excluded did not have.
The rest of the mechanism is older. The Gurus rejected caste as a religious principle, and the langar — where everyone eats together, seated on the floor — was the institution built to embody it. Part 8 described its founding.
The common claim is that Sikhism abolished caste. The evidence in front of us is that Punjab has India's largest Dalit population share and one of its sharpest land gaps, along with separate gurdwaras maintained by Dalit Sikh communities in many villages, and separate cremation grounds in some.
People usually reconcile these by saying the doctrine failed. But that reconciliation rests on an assumption worth pulling out: that a religion which rejects a hierarchy in worship is thereby attacking the hierarchy itself.
Caste in Punjab was never mainly a ritual ranking administered by priests — which is exactly why the Brahmin presence was weak. It was a distribution of land and of the labour attached to it. The langar attacks ritual separation at the meal. It says nothing about who owns the field the grain came from.
So the honest finding is not that the doctrine failed. It is that the doctrine and the hierarchy were never operating on the same thing. Equality at the meal was achieved and is real; it is one of the most successful institutions of its kind anywhere. Equality in land was not attempted.
This has a wider use. Whenever you are told that a belief system eliminated an inequality, ask which mechanism the inequality actually ran on, and whether the belief system ever touched it.
The 1961 law reserving a third of village common land for auction to Scheduled Castes is one of the few instruments aimed at the land question, and enforcing it has produced repeated village-level conflict — most visibly in the Sangrur district campaigns of recent years, where Dalit labourers organised to claim land they were legally entitled to lease.
9.4 — The protest#
In September 2020 the Indian Parliament passed three laws changing how agricultural produce could be bought and sold, loosening the role of the regulated market yards and enabling contracts directly with private buyers.
The government's case was that this freed farmers from a restrictive system and would raise incomes through competition. Farmers in Punjab read it as the first step towards ending assured procurement at a support price.
Chapter Six explains why the reaction was so much stronger in Punjab than elsewhere. In most of India, assured procurement barely operates, so removing it changes little. In Punjab it is the entire foundation of the cropping system — the thing that makes wheat and rice worth growing on a shrinking margin.
The protest that followed was one of the largest in modern history. Farmers camped at the borders of Delhi for over a year, through a winter and a pandemic. The laws were repealed in November 2021.
Both sides of the farm-laws argument shared a premise, and neither examined it: that the present cropping system is something to be either defended or opened to competition.
Neither position asked whether wheat and rice on this scale can continue at all. The government's case assumed private buyers would arrive for the same crops. The farmers' case assumed the state would keep buying them.
Chapter Seven contains the reason both may be wrong. The water is going. A system that cannot be sustained does not need to be defended or liberalised. It needs to be changed into something else — and the argument that consumed the country for a year was not about that.
Punjab went from India's richest state to the middle of the table — not by shrinking, but by growing slowly while manufacturing and services grew elsewhere. Agriculture cannot keep doubling.
Debt, opioid dependence and emigration are real and connected. Be sceptical of the enormous round numbers quoted about drugs; the honest figures describe a serious health problem, not a lost generation.
Punjab has India's highest share of Dalits and one of its widest land gaps. The langar achieved equality at the meal. Nothing ever attempted equality in the field, because caste here ran on land, not ritual.
The farm-laws protest was so much fiercer in Punjab because assured procurement is the whole foundation here. Both sides argued about who should buy the crop. Neither asked whether the water exists to keep growing it.
10An Honest List of What We Do Not Know
Every part of this series ends here. Two lists: what is genuinely unknown, and what is solid enough to build on.
This part has more in the first list than any other in the series, and the reason is not distance in time. It is the opposite. These events are recent enough that the people who could have counted chose not to, and are still alive to be asked.
10.1 — Genuinely unknown#
How many people died in the partition of Punjab. Estimates run from about 200,000 to two million. The gap is a difference of method — documented deaths versus demographic shortfall — not of data quality. Neither government investigated.
How many women were abducted, and what happened to most of them. The working figures of roughly 50,000 and 33,000 were administrative estimates for a recovery operation, not counts. What happened to women who were never recovered, or who were recovered against their wishes, is largely unrecorded.
Whether the Gurdaspur award was politically directed. The meetings happened; their content is unknown; Radcliffe's working papers were destroyed.
What happened to non-claimant refugees. Resettlement records everyone who filed a land claim. The landless, the artisans and the urban poor are largely invisible in the archive, and no systematic study has traced them.
How many people died in Operation Blue Star. No independent count exists. The press was excluded and cremations were carried out in bulk without identification.
The full scale of disappearances during the militancy years. Over two thousand unidentified police cremations were confirmed in Amritsar district. No equivalent examination was made in the rest of Punjab, so the statewide figure is not merely disputed — it has never been sought.
How much groundwater is left. The direction is certain and the rate is roughly known. The total remaining volume, and therefore how long Punjab has, is not, because aquifer thickness and yield vary and have not been mapped in enough detail.
Whether the militancy was ended by policing, by exhaustion, or by its own conduct. All three are argued by serious people and no public accounting has been made.
10.2 — Solid#
The sequence of 1947. The award's date, the districts affected, and the disbanding of the Boundary Force on 1 September are all documented and uncontested.
That the violence was organised. Official reports, army records and testimony from every side agree on armed bands, transport, lists and targeting. This is not a matter of interpretation.
That the violence began before partition was announced. The Rawalpindi killings of March 1947 predate the 3 June plan.
The scale of the movement. Roughly ten million people crossed in Punjab. Camp registers, ration records and resettlement claims all support this independently.
The design and effect of the resettlement scheme. The standard acre and the graded cut are fully documented, and the redistributive effect is visible in the allotment records.
That census language figures in 1951 and 1961 recorded declarations, not speech. The campaigns urging Punjabi Hindus to return Hindi are documented in the press of the period.
The results of the Green Revolution. Yields, procurement shares and income data are well recorded and agree.
That Punjab's water table is falling and why. Well measurements, pump-depth expenditure and satellite gravity data agree, and the last of these has no stake in the argument.
That November 1984 was organised and that police did not intervene. Established by multiple official inquiries, whatever word is chosen for it.
That Punjab has India's highest Dalit population share and a very large gap between that share and land ownership. Census and agricultural census data, both routine, both uncontested.
The largest gaps in this part are counts of the dead — in 1947, in June 1984, and during the militancy years. In each case the counting was not attempted, and in each case that decision was made by people who had the power to count.
What is solid is mostly structure: what was built, what was allocated, what was grown, what is under the ground. Institutions leave records even when they do not want to.
When someone gives you a confident total for any of the killings in this part, they are telling you which side they are on, not what happened.
Sources & further reading — Part 13
Timeline
Seventy-nine years, in the order they happened.
| When | What |
|---|---|
| March 1947 | Mass killing in the Rawalpindi division — before partition is announced. |
| 3 June 1947 | Mountbatten announces partition and the mechanism for dividing Punjab. |
| 23 June 1947 | The Punjab Assembly, voting in two sections, chooses division. |
| 8 July 1947 | Radcliffe arrives in India. He has about five weeks. |
| 14–15 Aug 1947 | Independence. The border is not published. |
| 17 Aug 1947 | The award is made public. |
| 1 Sept 1947 | The Punjab Boundary Force is disbanded as the violence peaks. |
| 1947–48 | Roughly ten million people cross Punjab in both directions. |
| April 1948 | India briefly stops canal water flowing into Pakistan. |
| 1948 | PEPSU formed from eight princely states. |
| 1949 | Abducted Persons (Recovery and Restoration) Act. |
| 1951, 1961 | Censuses in which large numbers of Punjabi Hindus return Hindi as mother tongue. |
| 1952 onward | Chandigarh built to Le Corbusier's plan. |
| 1955 | The States Reorganisation Commission refuses the Punjabi Suba. |
| 1956 | PEPSU merged into Punjab. |
| 1960 | Indus Waters Treaty: three eastern rivers to India, three western to Pakistan. |
| 1 Nov 1966 | Punjab divided into Punjab, Haryana and hill districts. Chandigarh and river waters left unresolved. |
| 1966–68 | High-yielding wheat spreads across Punjab. The Green Revolution begins in earnest. |
| 1973, 1978 | Anandpur Sahib Resolution adopted, then revised. |
| 1981–82 | SYL canal agreed and begun. Dharam Yudh Morcha launched. |
| Oct 1983 | President's rule imposed in Punjab. |
| 1–8 June 1984 | Operation Blue Star. The Akal Takht destroyed. |
| 31 Oct 1984 | Indira Gandhi assassinated. Organised killing of Sikhs follows in Delhi and elsewhere. |
| July–Aug 1985 | The Rajiv–Longowal accord. Longowal assassinated weeks later. Chandigarh not transferred. |
| 1984–1995 | Militancy and counter-insurgency. Mass killings by armed groups; mass disappearances by the state. |
| 1995 | Jaswant Singh Khalra abducted and killed after documenting secret cremations. |
| 1997 | Normal politics resumes. Free power to agriculture follows. |
| 2004, 2016 | Punjab terminates its water agreements; the Supreme Court holds this unconstitutional. The canal remains unbuilt. |
| 2011 | Census records Punjab's Scheduled Caste share at about 32 per cent, the highest in India. |
| 2020–21 | The farm laws, the year-long protest, and the repeal. |
| Now | Most of Punjab's groundwater blocks are classed as over-exploited. |
| Part 14 | Part 14 starts here — on the other side of the border, and in the countries Punjabis have moved to since. |
Glossary
Every hard word used in this part, in plain English.
| Word | What it means |
|---|---|
| Akal Takht | "Throne of the Timeless One." The seat of temporal authority in Sikhism, facing the Golden Temple across the pool. Destroyed in June 1984 and rebuilt. |
| Aquifer | A layer of underground rock or sand holding water between its grains. A very deep sponge under the fields. |
| Graded cut | The rule that reduced every land claim, but reduced large claims by a far bigger proportion than small ones. |
| Gurmukhi | The script Punjabi is written in in India; the script of the Guru Granth Sahib. |
| HYV | High-yielding variety. A crop bred for far more grain per plant — for wheat, by making the plant shorter so it does not fall over under fertiliser. |
| Jatha | A band or squad. Used for unarmed Akali volunteers in the 1920s and for armed bands in 1947. |
| Kafila | A caravan. In 1947, a refugee column moving on foot, sometimes hundreds of thousands strong. |
| MSP | Minimum Support Price. A price announced before sowing at which the government undertakes to buy the crop. |
| Muhajir | Migrant. Used in Pakistan for those who came from India, with a religious echo India's word "refugee" does not carry. |
| Over-exploited | The official classification for an area where more groundwater is pumped out each year than soaks back in. |
| PEPSU | Patiala and East Punjab States Union, 1948–1956. Eight princely states merged into one unit, later absorbed into Punjab. |
| Princely state | Territory ruled by an Indian prince under British paramountcy rather than governed directly. |
| Procurement | The government actually buying grain at the support price, for the national food distribution system. |
| Recharge | Water soaking down from rain and rivers to refill an aquifer. |
| Riparian principle | The idea that the states a river flows through have first claim on its water. |
| Standard acre | Land measured by what it yields rather than how big it is. The unit that made resettlement possible. |
| Suba | Province. "Punjabi Suba" was the demand for a Punjabi-language state. |
| SYL | Sutlej-Yamuna Link. The canal agreed in 1981 to carry water to Haryana. Begun in 1982, never completed. |
| Tehsil | An administrative unit below a district. Which unit you count majorities in changes where a border falls. |
| Tubewell | A pumped well drawing water from underground. Punjab has over a million. |